introduction - harvard university...introduction 3 with information about how the world around him...

Introduction

There are so many people recording stories these days. For the most part they are all about ghosts and gods and mutations and monsters, stuff that is absurd and ridiculous; or else they are entertaining and humorous, serving as a source of laughter and merriment. Beyond these two types, some criticize past affairs—all these slander former worthies, causing the multi-tudes to take them as a subject for gossip. This is the pervasive ailment of the present age.

近日著小説者多矣。大率皆鬼神變怪荒唐誕妄之事。不然則滑稽詼諧以為笑樂之資。離此二者。或强言故事。則皆詆訾前賢。使悠悠者以為口實。此近世之通病也。

—Lu Xisheng 陸希聲

Ninth-century China witnessed a fad for telling and collecting stories about remarkable events. Accounts of political maneuvering, romantic entanglements, encounters with ghosts and were-beasts, and other strange doings circulated both orally and in writing within the community of elite men who were also the stories’ chief protagonists. Today we can find the traces of this fad in the written tales recorded by members of this commu-nity, though the tales that survive are only a fraction of the number in circulation at the time. They range from short accounts that occupy several lines in a modern printed edition (a paragraph or two in English transla-tion) to longer and more elaborate narratives covering several pages. Like poetry from the same period, the written tales were both social and textual objects. They were influenced by stories exchanged at gatherings intimate and public, but themselves were written texts, recorded for reading and for circulation in written form. Their language is terse literary Chinese, although the style of individual tales may lie anywhere on a spectrum from unadorned narrative to ornate and densely allusive prose. In scope they tend to focus on a limited incident in the life of one protagonist, most often an elite man—highly educated, trained to seek and serve in office in the

Epigraph: Lu Xisheng, “Beihu lu xu” 北戶錄序 (Preface to Beihu lu), dated to ca. 874 (Duan Gonglu, Beihu lu, “Xu,” 1; for dating, see Fu Xuancong, Tang Wudai wenxue, 3:625).

Allen Final Pages.indd 1 7/25/14 10:29 AM

2 Introduction

state bureaucracy, and steeped in the literary tradition—and to assume the limited viewpoint of that protagonist.1 They deal with matters of private experience rather than public record, their events knowable only by an eyewitness and not subject to public verification. Lu Xisheng’s disapprov-ing remarks about this “ailment,” recorded near the end of the period that this study covers, hint at the appeal that such tales had for their audiences in his own age and beyond. These “absurd and ridiculous,” “entertaining and humorous,” and “gossipy” tales are the subject of this book.

Lu Xisheng’s categories aptly capture the range of topics found in the tales, despite his dismissive tone. Humorous tales form only a minority; his other two groupings encompass the majority of surviving material. What Lu Xisheng describes as accounts of “ghosts and gods and mutations and monsters” can be more broadly characterized as lore whose chief concern is the creatures and spaces that lurk just beyond ordinary perception and may show themselves to a man in the right (or wrong) place at the right time. A large proportion of surviving tales from the eighth and ninth centuries center on incidents that transgress the boundaries of ordinary experience and touch on the occult, in varied manifestations. A man may find himself turning into a tiger or realize that his wife is actually a were-fox. He may unexpectedly be dragged off to hell to do some underworld official a favor or to plead for his own life; or he may find himself feasting with gods or transcendents beyond the mortal realm. He may witness the wondrous deeds of a man (or woman) endowed with mysterious powers or unwit-tingly be drawn into commerce with a ghost. Such tales describe the ordi-nary person’s encounter with something extraordinary, experiences that are anomalous to the course of the protagonist’s everyday life. The same item of lore, more or less, might be found set in different locales and times with different actors and details, much like the urban or modern legends studied by contemporary folklorists, with the important distinction that the actors in Tang tales are typically named.2 In aggregate these tales present the reader

1. That is, the events recounted are focalized, exclusively or almost exclusively, through the eyes of the protagonist experiencing them (for the notion of “focalization,” see Genette, Narrative Discourse, 185–94; and Bal, Narratology, 142–61).2. For discussions of the characteristics and modes of dissemination of urban legends, see Brunvand, Vanishing Hitchhiker, especially the general discussion in the first chapter (1–17); and DiFonzo and Bordia, “Rumor,” 28–30. As DiFonzo and Bordia note, the same urban legends (which may or may not be set in urban locales) reappear across different


Introduction 3

with information about how the world around him works: foxes may take human guise; ghosts of the dead may still feel the desires of living humans; and, through diligent self-cultivation, men and women may achieve powers beyond the lot of the ordinary person. Lu Xisheng’s reference to tales that become “a subject for gossip” points to tales that relay particular information about particular historical people, a second major type in the tale corpus. Although many works in this cate-gory show their subjects in a good light rather than engaging in the slander that Lu Xisheng deplores, tales centered on gossip are interesting in part or in whole because of whom they are about—often public figures who would have been well known in their day. These tales also convey information, but instead of (or in addition to) presenting a model of experience that might apply to whosoever happened to be in a certain place at a certain time, they reveal what happened to a specific individual, a person presumably known to the tale’s audience through reputation if not personal acquain-tance. Gossipy tales frequently concern prominent officials or literary figures whose names and activities appear also in other historical sources; they reveal the otherwise private lives and hidden operations of men in the upper echelons of government or society. Their utility to the reader thus lies largely in their value as historical information, describing what has happened in the past rather than presenting a model for what might happen at any time to anyone (though, as with the lessons of any history, readers may hope to find guidance for their own behavior in them). Recognizing the name of the person or people such a tale is about changes the reader’s apprehension of its information: it is not simply a sequence of interesting events, but interesting events that happened to a particular person. Gossip and lore are not mutually exclusive within tales. Many tales combine both elements: an anecdote about the eighth-century minister Li Gao 李暠 (d. 740) frightening a demon disguised as a beautiful woman tells us something about demons (they can take the form of beautiful women, but an upright man may force them to show their true forms) and about Li Gao (he was so upright that demons quailed at his presence).3 Tales that read as lore-based today may also have been recognized as gossip in

places and times, and frequently contain a moral or a warning. Brunvand emphasizes that urban legends are invariably presented as true stories, though the actors within them are rarely named (4).3. “Li Gao,” in TPGJ, 329.2615 (citing Guangyi ji 廣異記).


4 Introduction

their own time and locale, before the named protagonist had been forgot-ten beyond his role as a character in a tale. To readers who know or know of their protagonists, the information these tales impart about the men named within them is as important as their revelations about the existence and habits of demons and other beings on the margins of humanity. The thread that holds the diverse strands of this corpus together is their focus on private experience and private information. These tales fix in writing narratives whose content excludes them from the formal public record, whether because they touch on the occult, because they concern matters that occur behind closed doors, or because their subjects are simply not important enough to be remembered there—often all three.4 To say that Tang tales take lore and gossip as their subject matter is not to say that they are no more than hearsay recorded in writing, however. Though the majority of tales likely drew on received information and events, the uses to which that information and those events were put varied considerably from tale to tale and from recorder to recorder. In the hands of some writers they maintain the flavor of hearsay, recorded to preserve the memory of a remarkable event but little elaborated beyond that; but for another writer the same incident might serve as the starting point for a carefully crafted and richly imagined story.

Tales and Fiction

It is because of the carefully crafted and richly imagined tales within the corpus that Tang tales came to new prominence in the early twenti-

4. In this they fit easily with Andrew Plaks’s argument that the distinction between “fiction” and “history” in the Chinese tradition lies largely in the subject matter with which a work engages. He writes, “For want of a differential truth-function or sharply contrasted narrative form, the major observable difference that conspicuously separates the two branches of Chinese narrative is the simple fact that historiography (and historical fiction) deals primarily with affairs of state and public life—military, political, diplomatic, court-related—while fiction takes up the slack to cover the more individualized and inti-mate details of the private lives of figures of varying roles or statures” (“Towards a Critical Theory,” 318; emphasis in the original). Plaks’s discussion focuses on vernacular fiction and references classical language tales only briefly, but his distinction between public and private life is a useful one for considering the place of Tang tales in the written record of the time.


Introduction 5

eth century, when China’s intellectuals were urgently concerned with the question of how to transform China into a modern nation that could take its place on an increasingly global stage. Several writers, the essayist and short-story writer Lu Xun 魯迅 the most influential among them, began reexamining the Chinese literary tradition in light of categories found in (Western) histories of Western literature—a move that was part of an attempt to give China’s “national culture” a place in world history.5 The Tang was identified as the era when fiction emerged in China, evolving out of shorter, rougher tales of the previous Six Dynasties period that dealt with much of the same subject matter but treated it less imaginatively. In his Zhongguo xiaoshuo shilüe 中國小說史略 (Brief history of Chinese fiction), initially published in 1923, Lu Xun posited that the Tang was the first period in which Chinese writers intentionally produced fictional tales, writing that “narration became more subtle, wording more lovely—the marks of progress are quite clear when compared with the coarse accounts and broad outlines of [tales of] the Six Dynasties. And what is particularly evident is that it is at this time that [writers] first deliberately wrote fiction (xiaoshuo)” 敘述婉轉，文辭華艷，與六朝之粗陳梗概者郊，演進之跡甚明，而尤顯者乃在是時則始有意為小說.6

The crucial term Lu Xun uses to describe his subject matter is xiaoshuo 小說 (literally, “small talk”), an old bibliographic category that in the late nineteenth and early twentieth centuries had begun to take on the new meaning of “fiction” through the influence that Western ideas about Western literary history had on both Japanese and Chinese scholars.7 To say that Tang tales were xiaoshuo was not in itself new in the twentieth century. Lu Xisheng had used the term in the remarks quoted in the epigraph to this introduction. Likewise Sun Guangxian 孫光憲 (900–68) used it to describe a collection of tales when he called the Tang collection Xuanshi zhi 宣室志, attributed to Zhang Du 張讀,

5. See Glen Dudbridge’s succinct but incisive discussion in “Question of Classification,” 196–97.6. Lu Xun, Zhongguo xiaoshuo shilüe, 44.7. See Dudbridge, “Question of Classification,” 196–97; and Zeitlin, “Xiaoshuo,” 249–59. For various readings of the changing Japanese use of the term 小說 (Japanese shōsetsu) and its adoption as the equivalent of “fiction” in the nineteenth century, see Zeitlin, “Xiaoshuo,” 259–60; Zwicker, “Long Nineteenth Century,” especially 570–74; and Suzuki, Narrating the Self, 16–23.


6 Introduction

“Vice Director Zhang Du’s xiaoshuo” 張讀侍郎小說.8 Many centuries later, the Ming dynasty scholar Hu Yinglin 胡應麟 also used xiaoshuo in a passage in which—like Lu Xun—he praises the inventiveness of Tang tales, contrasting their “purposeful pursuit of novelty” 作意好奇 with the simpler “recording of falsities” 傳錄舛訛 in material from the earlier Six Dynasties period.9 But the valence of the word xiaoshuo had shifted a number of times over the centuries since its initial introduction as a bibliographic category in the first century, when the historian Ban Gu 班固 defined the category xiaoshuo as “chit-chat of the streets and alleyways, created by those holding conversations along the roadside” 街談巷語，道聽塗說者之所造也.10 In Tang texts the term is most often applied to shorter, more gossipy items than the material Lu Xun treats as “fiction” (xiaoshuo) in his Zhongguo xiaoshuo shilüe; the Southern Song (?) Zuiweng tanlu 醉翁談錄 (Record of an old drunkard’s conversations), attributed to Luo Ye 羅燁, uses it to refer to professional oral story-telling; and, with the growth of the vernacular short story and novel in the late imperial period, it was applied to that material as well.11 The modern usage of xiaoshuo as an equivalent, more or less, for the Western notion of fiction imposed onto the Tang corpus (Western) romantic associations with authorship, creativity, and originality of conception that fit only uneasily with the material itself. Lu Xun’s history makes clear that he was aware of the dissonance between the traditional bibliographic category xiaoshuo and the Western concept of “fiction” that was increasingly being associated with xiaoshuo,

8. Wang Meng’ou, Tangren xiaoshuo jiaoshi, 1:i; Sun Guangxian, Beimeng suoyan, 7.152.9. Hu Yinglin, Shaoshi shanfang bicong, 36.486. Lu Xun himself quotes these phrases at the beginning of the first of his three chapters on the Tang (Zhongguo xiaoshuo shilüe, 44).10. Ban Gu, Han shu, 30.1745, translated in Zeitlin, “Xiaoshuo,” 249.11. For a succinct but more thorough account of the history of the use of xiaoshuo than I can give here, see Zeitlin, “Xiaoshuo,” 249–59; Ming Dong Gu covers similar ground with a different focus in Chinese Theories of Fiction, 22–25. Examples of the Tang usage include Liu Su’s 劉餗 preface to his Sui Tang jiahua 隋唐嘉話, which he describes as “bits of small talk” 小說之末; Li Zhao 李肇 echoes this description of Liu Su’s collection of short gossipy notes in the preface to his Guoshi bu 國史補 (Sui Tang jiahua, 1; Guoshi bu, 3). This sort of material is often pointedly excluded from the category of xiaoshuo proper or separated off as biji xiaoshuo 筆記小說 (note-form xiaoshuo) in scholarship today (for a discussion of biji xiaoshuo and why it is not classed as xiaoshuo in most modern scholarship, see Li Jianguo’s discussion in Xulu, 1:3). For Luo Ye’s usage, see Zuiweng tanlu, 3; the work is believed to date to the late Southern Song.


Introduction 7

if only because his discussion of early xiaoshuo acknowledges its status as gossip and hearsay rather than “fiction” per se. But his reference to “deliberately writing fiction (xiaoshuo)” as well as to “deliberate creation” 有意之創造 in describing the Tang material suggests that he did intend to associate Tang tales with fiction (while acknowledging that not all xiaoshuo was “fiction”).12 Lu Xun’s comments have been echoed and quoted ever since, and the notion that Tang tales are fiction has become a standard assumption in much of the scholarly discourse concerned with these tales.13

Although the association of the Tang material with fiction had more to do with the goals of early-twentieth-century intellectuals than with a comprehensive assessment of the corpus itself, once the link was estab-lished, Tang tales took on a new significance for literary historians because they became “the beginnings of fiction in China.” As Glen Dudbridge writes, “Fiction/xiaoshuo became canonical in the twentieth century, [and] historians of Chinese literature have ever since been content to refocus their categories accordingly.”14 The identification of Tang tales with fiction is made in contrast with narratives that, though they may also be narra-tives of untrue events, are presented as documentary in intent: that is, the events recounted may not have actually happened, but the recorder proba-bly believes that they did and recounts them with the expectation that his audience will take them at face value—Hu Yinglin’s “recording of falsities.” Fiction and documentation are thus used to mark opposing poles char-acterized respectively also by imagination and reportage, complexity and simplicity, literariness and plain writing. In this scheme the finest tales are those seen to depart furthest from reportage, in which the writer—the author—gives free rein to his imagination and inventiveness as he “deliber-ately” (as Lu Xun puts it) fabricates his story. The notion that the fabrica-tion of one’s subject matter is the mark of true literariness in tales is found

12. Lu Xun, Zhongguo xiaoshuo shilüe, 44.13. See, for example, the entries “Chih-kuai ” and “Ch’uan-ch’i ” in Nienhauser, Indiana Companion, 280–84 and 356–60; and Sheldon H. Lu’s chapter “Reading T’ang Fiction,” in From Historicity to Fictionality, 93–128. Mair discusses Tang tales at length as “fiction” (Columbia History, 579–94). Given the wider valence of xiaoshuo in Chinese, Chinese-language discussions remain more ambiguous, but (to give just one example) Wu Zhida’s use of “imagination” 想像 and “fabrication” 虛構 to describe Tang xiaoshuo indicates the same assumptions about fictionality (Tangren chuanqi, 12).14. Dudbridge, “Question of Classification,” 197.


8 Introduction

even in the work of scholars thoroughly conversant with the full corpus, with all its repetition and overlap.15

Though it is true that a number of tales in the Tang corpus have elements that we may associate with fiction—interesting plots, fine writ-ing, detailed description, suspenseful pacing, attention to characters’ psychological state—the vision of these tales that emerges when we put aside the straightforward fiction/not-fiction rubric is much more interest-ing and complex. Considering Tang tales as fiction restricts us to a small proportion of the extant material, which, though superficially it may seem to admit the label of “fiction,” limits our understanding of both the select “fictional” tales and the larger tale corpus of which they are a part.16 The counterargument to this point is that it is the select tales that are fiction, standing out in a field of inferior works. But this argument is problematic because the continuities between the “fictional” tales and the rest are so great that it makes little sense to separate them: they recount the same stories, share the same themes, and are written in the same language, often by the same writer and within the same collection. More fundamentally, there is no evidence that the writers of these tales conceived of them in terms of a “making” of something new that is associ-

15. Thus Li Jianguo, the foremost contemporary Chinese scholar to work on this material, acknowledges the indebtedness of many tales to hearsay (Xulu, 1:17) but at points appears to regard this fact somewhat regretfully. In describing the mid-ninth-century collection Ganze yao 甘澤謠, attributed to Yuan Jiao 袁郊, for example, he writes: “In some cases [Yuan Jiao] uses source material, yet he subtly expresses his thoughts, singularly wielding his ax, forging his own mighty words, completely novel in conception” (Xulu, 2:806). The “yet” 然 implies an opposition between “using source material” and “novelty of conception.” Li Mei’s 李玫 Zuanyi ji 纂異記, in contrast, receives Li Jianguo’s praise as “an exquisite work of Tang storytelling. . . . All the pieces come from personal invention, not relying on hearsay” (Xulu, 2:714). The tales in Zuanyi ji are, from a narrative standpoint, also among the least interesting to survive, tending to use simple narratives as a framework into which to put poems or other types of quoted material.16. Among early but still-influential studies of Tang tales, Liu Kairong ostensibly takes the thirty-two Tang tales in Lu Xun’s anthology Tang Song chuanqi 唐宋傳奇 as the basis for her study in Tangdai xiaoshuo yanjiu (first published 1947), but, as Sarah Yim has pointed out, Liu actually only deals with ten tales (Yim, “Structure, Theme, and Narrator,” 23, n. 28). Zhu Xiuxia’s Tangdai chuanqi yanjiu uses thirty-nine tales. Yim’s own dissertation, one of the first studies of the material in English, is based on sixty-eight tales. A number of more recent studies impose similar restrictions, as for example Wu Zhida’s Tangren chuanqi, which focuses on a relatively small selection of anthology pieces.


Introduction 9

ated with fiction;17 rather, the process of composition, in the few instances when we can find one articulated, is that of collecting material and filling in the gaps in an incomplete record. Thus, in a rare surviving fragment of a preface to a Tang tale collection, Lu Zhao 廬肇 describes the method by which he compiled his collection Yishi 逸史 as one in which he “gathered up the unusual things that I had heard of and read about . . . for all of them just picking out the substance, and filling in the gaps” 集聞見之異者 . . . 皆摭其實補其缺而已.18 My argument is not that no tale was ever invented by its writer as he recorded it: for the majority of tales we have no way to tell how much a writer relied on sources and how much he invented himself. But the dominant rhetoric is that of supplementation rather than of creation, and when we turn to the tales themselves, we see that in many instances this is precisely what writers did: what distinguishes more and less finely crafted tales is the degree of elaboration with which the gaps get filled. The debate on fictionality touches on the terminology used to label these tales as well. Another legacy of Lu Xun’s Zhongguo xiaoshuo shilüe that became a truism of twentieth-century scholarship is the division of the tales into two subcategories, chuanqi 傳奇 (transmitting the marvel-ous) and zhiguai 志怪 (recording the weird). As the terms suggest, the two subcategories are conceived of as closely related, both comprising accounts of extraordinary events; but the longer, more artfully wrought tales are chuanqi, whereas the shorter and simpler accounts are zhiguai. Lu Xun identifies chuanqi as the Tang successor to Six Dynasties zhiguai. In his account the relation between the two forms is sequential: chuanqi grew from and improved upon the earlier zhiguai. He does not address whether or not zhiguai continued to be written in the Tang.19 More recent schol-arship expands the scope of the material addressed and discusses both “Tang chuanqi ” and “Tang zhiguai,” maintaining Lu Xun’s terminology but allowing zhiguai to continue into the Tang even as chuanqi emerged, thus accounting for the short tales that make up the majority of the Tang corpus.

17. Cf. the oft-cited origin of fiction in the Latin fingere, “to fashion or form” (Compact Oxford English Dictionary, 584).18. SF 24.21b (3.1740). The preface survives only in SF, which leads Wang Rutao to doubt its authenticity (Quan Tang xiaoshuo, 1:813) and Li Jianguo to rejoice over its chance pres-ervation (Xulu, 2:692). See also Shen Yazhi’s 沈亞之 remarks about “filling in” 補 words in the recording of his “Explanation of the Plaint from the Xiang” 湘中怨解 (Shen Yazhi’s tale is found in SXXJ, 2.1a–2a; and WYYH, 358.835c–36a).19. Lu Xun, Zhongguo xiaoshuo shilüe, 44–45.


10 Introduction

The distinction between the two approximates that between fiction and not-fiction, and encounters the same difficulties: there is no way to draw a clear line between the two subcategories.20 Nor—as with xiaoshuo—was either chuanqi or zhiguai used as a generic label in the centuries when these tales were recorded.21 Tang writers most often referred to tales simply as “accounts” (zhuan 傳) or “records” ( ji 記), terms that were used to apply to a much wider range of materials than chuanqi or zhiguai or even xiaoshuo.22 Thus, as with the fiction/not-fiction divide, the terms zhiguai and chuanqi are more useful as a reminder of the spectrum that is encompassed by the tale corpus than as generic labels. It is because of the lack of an apposite established nomenclature for this corpus that I use “tale” to refer to the material I treat in this book, whether simple or elaborated, short or long,

20. A number of scholars who nonetheless maintain the categories admit as much. Li Zongwei suggests that scholars fall back on the more general term xiaoshuo because the scope of chuanqi has not been clearly delineated (a problem he seeks to rectify; Tangren chuanqi, 8–11). Li Jianguo concedes that in many cases determining which category a tale belongs in requires approximate and instinctive distinctions, and the imposition of artifi-cial rules about length (Xulu, 1:5). Wang Meng’ou, however, argues against using the term chuanqi because of its anachronism (he does not discuss zhiguai; Tangren xiaoshuo jiaoshi, 1:i–ii). Dudbridge goes even further and suggests, persuasively, that pigeonholing tales through the use of terms such as chuanqi (and zhiguai) is in fact deleterious to our attempts to understand the material (“Question of Classification,” 197–98 and 207–8).21. See Xulu for an overview of the usages of both terms; Li Jianguo writes that to his knowledge the first use of the term chuanqi to refer to a category of tale is found in Xie Caibo’s 謝采伯 preface to his Mizhai biji 密齋筆記 (1241), in which he refers to “types [of writing] like chuanqi and zhiguai ” 傳奇志怪之流 (Xulu, 1:6–9). Rania Huntington notes that the first critic to use the terms as “clear subdivisions of xiaoshuo” was Hu Yinglin but that they only became widely accepted after Lu Xun; see also her insightful discussion of the problematic relationship among zhiguai, chuanqi, and fiction (Alien Kind, 15–20).22. Since contemporaneous commentary on the tale corpus is virtually nonexistent, it is difficult to identify a consistent terminology with which tales are described. Tale writers themselves sometimes make reference to “making an account” (傳 or 作傳) as a description of their work in recording a tale. Thus, in “Miss Ren” 任氏, Shen Jiji 沈旣濟 says that his friends “asked Jiji to make an account of it” 請旣濟傳之 (TPGJ, 452.3697); in “Xie Xiao’e” 謝小娥傳, Li Gongzuo 李公佐 speaks of “writing an account” 作傳 of Xie Xiao’e’s story (TPGJ, 491.4032). Both ji 記 and zhuan 傳, as well as lu 錄 (also “record”), are found frequently in the titles of individually circulating tales. But more generalized categorizations are rare and tend to cite earlier material. For example, the seventh-century Sui shu 隋書 classes similar material from previous eras under “miscellaneous accounts” 雜傳 (33.974–82), and the histo-rian and historiographer Liu Zhiji 劉知幾 (661–721) refers to “miscellaneous records” 雜記 (Shitong tongshi, 10.249). Li Jianguo points out that Northern Song (960–1127) writers also often referred to Tang tales as “accounts and records” 傳記 (Xulu, 7).


Introduction 11

documentary or inventive—while acknowledging that the absence of defin-itive Tang terminology or an explicit generic conception from that period means that the corpus itself must remain somewhat loosely defined.

Tales and Collections

More significant when these tales were first recorded than the chuanqi- zhiguai divide was that between individually circulating tales and tale collec-tions.23 Most readers today—and for the past several centuries—encounter Tang tales in anthologies or compendia, which place side by side items from different sources and by different writers. Anthologies promise to select the finest tales; compendia such as Taiping guangji 太平廣記 (on which more below) claim comprehensiveness rather than selectivity but rearrange tales by topic. Both obscure the contexts in which the tales were originally recorded and circulated, and in so doing they also erase the distinction between collec-tions of tales and tales written to circulate independently. The majority of surviving tales originally circulated within collec-tions of tales compiled by a single writer.24 Tang tale collections drew on and continued a tradition of gathering accounts of strange events—what Robert Ford Campany has termed “anomaly accounts”—that emerged in the fourth century (these are the early zhiguai from which chuanqi are said to have evolved).25 As Campany has established, the chief motivation

23. Some scholars equate tales in collections with zhiguai and individual tales with chuanqi, but, as Li Zongwei points out, such a restriction excludes many fine tales from the realm of chuanqi; most discussions allow for chuanqi both outside of and within collections (Tangren chuanqi, 9–10).24. Li Jianguo’s Xulu covers 85 collections and 112 individual tales that he dates to the Tang; the collections included anywhere between around a dozen and a few hundred tales.25. Campany writes, “From the fourth century . . . a new object of discourse comes into focus, the anomalous event per se, and is dealt with within a new textual form—a sheer list of discrete, disjointed narrative items not interconnected by common figures or by any overarching frame or single theme or topic (though sometimes arranged chronologically or topically). Along with this new textual form comes an unprecedented focus on ordi-nary human protagonists—mostly (but not all) members of the elite, to be sure, but also low- and middle-ranking members, as opposed to divine or legendary figures, rulers, or religious specialists” (Strange Writing, 100; emphasis in the original. For the term “anomaly accounts,” see also ix and 2–3).


12 Introduction

behind the compilation of these early tale collections was cosmographic: they attempt to chart and describe the world and its workings.26 By the mid-eighth century, the beginning of the period with which I am primar-ily concerned here, the scope of the tale collection had widened from the chiefly religious impetus directed at proving the existence of the occult or the efficacy of a particular type of practice (Daoist or Buddhist) that characterizes earlier tales to a more catholic interest in occult and unusual events of a variety of kinds.27 Whatever its motivation, however, a collec-tion typically represented a long-standing commitment to collecting tales about strange things over time. Collections might contain a dozen tales or less, or a few hundred; longer collections in particular might grow and evolve over time as a writer gradually added to his stock of material.28 Tales in many collections, especially from before the mid-ninth century, retain the flavor of recorded hearsay: they are short, often vivid but not greatly elaborated, and focused on evoking a single incident. A minority of tales—among them many of the most elaborate and best known today—were individually circulating tales, tales written to stand alone. Individual tales represent the fruit of a one-time impulse rather than a sustained fascination with strange events; a number are explicitly framed within the text as a writer’s response to an event that impressed him (and sometimes his friends) and that he is then moved to record.29

26. Ibid., 2–8.27. Thus, while Tang Lin’s 唐臨 seventh-century collection Mingbao ji 冥報記 is composed entirely of Buddhist miracle tales, the mid- to late-eighth-century collections Jiwen 紀聞 (attributed to Niu Su 牛肅) and Guangyi ji (attributed to Dai Fu 戴孚) are both more eclectic in their contents, including items of both Daoist and Buddhist flavor, and of neither. But nor did the doctrinally inspired collection disappear, as we see from Du Guangting’s 杜光庭 (850–933) early-tenth-century compilations of Daoist hagiographies. Wu Zhida also notes this shift away from narrowly religious motivations but frames it in terms of the distinction between zhiguai and chuanqi, writing, “The creation of zhiguai was fundamentally a religious activity 宗教活動, but the creation of chuanqi, in contrast, was fundamentally an aesthetic activity 審美活動” (Tangren chuanqi, 11).28. Li Jianguo posits that a number of collections were built over time, including Li Fuyan’s李復言 Xu Xuanguai lu 續玄怪錄, Duan Chengshi’s 段成式 Youyang zazu 酉陽雜俎, and Pei Xing’s 裴鉶 Chuanqi 傳奇 (Xulu, 2:694, 2:748, and 2:872).29. A number of writers experienced such a “one-time” impulse more than once, however: Li Gongzuo is credited as the writer of four surviving individually circulating tales: “The Governor of the Southern Branch” 南柯太守傳, “Old Lady Feng of Lujiang” 廬江馮媼傳, “The Ancient Classic of Mountains and Rivers” 古岳瀆經, and “Xie Xiao’e”—which ends with the explanation that “to know of goodness and not record it is not the intent of the


Introduction 13

Those that survive tend to be more complex and more artfully wrought as texts, and often were longer and more detailed than many of the stories found in collections (though collections themselves might include a mix of a few long, elaborated tales and many that were much shorter). As a group they share the collection’s focus on strange or unusual events, and the same origins in gossip and lore; but some individual tales eschew the occult entirely and focus on events whose chief strangeness lies in human behavior and human emotion. Though a few individually circulating tales are dated to the early years of the dynasty, the majority known today date from the period from the 780s through around 820. Though the distinction is harder to discern today when all tales are typically printed in conjunction with other tales, we must imagine that in the eighth and ninth centuries when these tales were first recorded and circulated, the difference between a solitary tale and a collection was signif-icant: the former presenting a single set of events for the reader’s consider-ation, to be read—even savored—in a single setting; the latter inviting the reader to finish one tale and read the next. A single tale was more easily copied and thus presumably more easily circulated, though perhaps as a shorter text also more easily lost; to copy an entire collection, especially a long one, required a commitment. Up through the first decades of the ninth century, the two types of tales appear to have been fairly distinct, with collections consisting almost exclusively of shorter tales and most of the more elaborated tales circulating as independent texts.30 Because this

Spring and Autumn Annals, thus I have composed this account to honor and praise her” 知善不錄。非春秋之義也。故作傳以旌美之 (TPGJ, 491.4032). Shen Yazhi is credited with five: “Strange Dreams” 異夢錄, “The Explanation of the Plaint from the Xiang,” “Feng Yan” 馮燕傳, “A Dream of Qin” 秦夢記, and “The Marvelous Response” 感異記 (attribu-tion tentative; see Xulu, 1:410–11).30. That at least some readers saw an affinity between them is shown by Gu Kuang’s 顧况 preface to his friend Dai Fu’s tale collection Guangyi ji, in which he names both early collections (including Gan Bao’s 干寶 Soushen ji 搜神記 and Tang Lin’s Mingbao ji) and two individual tales (“Wang Du’s ‘The Ancient Mirror’ ” 王度古鏡記 and “The Four Gentlemen of the Liang” 梁四公記) among Guangyi ji ’s antecedents. However, Gu Kuang’s list of antecedents casts a wide net; moreover, the two individual tales he mentions are both episodic in structure, carrying the flavor of many short tales linked together. Whether he would have considered the long individually circulating tale to be entirely of a kind with Guangyi ji is unclear. (For Dai Fu’s preface, see WYYH, 379.1749c–50a [the antecedents are listed on 1749c], as well as Dudbridge’s translation and commentary in Religious Experience and Lay Society, 18–45.)


14 Introduction

distinction appears to have been such a significant one in the period when the tales were written—and because of the considerable scope in stylis-tic range found in the tales and the absence of a consistent contemporary nomenclature—I refer to this material as a “corpus” rather than attempting to define the tale as a “genre.” But, by the middle decades of the ninth century, the scope of the tale collection had widened further as a few collection writers included more complex and carefully crafted tales within their collections. Different collec-tions elaborated their tales in different ways. Many of the pieces in Li Mei’s Zuanyi ji use simple narratives as frameworks for the insertion of other genres, chiefly poems chanted by the characters in the course of the events recounted. Tales in Yuan Jiao’s Ganze yao transform inherited stories in novel ways. Pei Xing’s Chuanqi also builds on inherited material in conven-tional plots, but its tales frequently introduce unusual twists, postponing expected resolutions and combining themes more often left separate. Like many individually circulating tales, these and some other late collections would seem to be intended for the reading pleasure of a literate audience who is familiar with the tale’s already-established conventions. But rather than consciously created fiction, these tales are more aptly characterized as (consciously) testing and manipulating the familiar themes and conven-tions of a corpus that remains rooted in gossip and lore.31

Roughly contemporaneous with Chuanqi, a man named Chen Han 陳翰 put together the only anthology of Tang tales known to have been compiled in the Tang, titling it Yiwen ji 異聞集 (Collection of strange hear-say). All the tales that Chen Han collected appear to have circulated inde-pendently up to that point rather than being part of prior single-author collections. By forming an anthology (a multiauthor collection) out of tales that had previously circulated individually, Chen Han linked the individ-ual tales, as a group, to the well-established tradition of the collection—an identification underscored by the title Yiwen ji, which is reminiscent of the titles of a number of other tale collections from the eighth and ninth centuries.32 Both the inclusion of more elaborate tales in collections and

31. My argument is similar to one made by Li Jianguo, who makes a case that it is a new “self-consciousness” 自覺 about tale writing that marks the difference between simpler tales (he uses zhiguai) and more complex (chuanqi). See Xulu, 1:30 and 1:81.32. Examples include Guangyi ji, “Wide-Ranging Records of the Strange”; Boyi zhi 博異志, “Accounts of Surpassing Strangeness”; Jiyi ji 集異記, “Records of Strange Things Collected”; and Duyi zhi 獨異志, “Accounts of the Uniquely Strange.”


Introduction 15

the compilation of Yiwen ji bridge the gap between the two types of tales, already linked by shared subject matter and shared plots, such that by the mid-ninth century the tale collection was conceived of very differently from the vision of the collection as a repository of unelaborated lore inherited from Six Dynasties compilations.

The State of the Material

Although Tang tales circulated in collections or as individual texts in their own day, very few collections, and even fewer individual tales, have come down to us today in their original form. No Tang era manuscripts of tales or collections survive; the earliest material trace of the tales to be preserved is a partial manuscript of the tale “Travels in the Land of Zhou and Qin” 周秦行紀 dated in a colophon to 935, found in the caves in the oasis town of Dunhuang when its trove of documents was rediscovered near the turn of the twentieth century.33 The Beijing Library has an incomplete and undated woodblock print edition of Li Fuyan’s collection Xu Xuanguai lu produced by a Southern Song (1127–1279) printer, the earliest edition of a collection to be preserved. A few collections and individual tales (the latter reprinted in a larger compendium) also survive in printed or manuscript editions from the Ming or Qing dynasties, but these represent only a frac-tion of the total surviving material (underscored by the fact that none of these editions is itself believed to be complete).34

33. The tale is found in the Dunhuang manuscript Pelliot 3741, which begins roughly a third of the way into the story; the earlier portion appears to have been torn off (Bibliothèque nationale de France, Manuscrits de Dunhuang, 27:217–18).34. “Travels in the Land of Zhou and Qin” and a few other individual tales (including “The White Ape” 白猿傳 and “The Curly-Bearded Stranger” 虬髯客傳) are reprinted in the sixteenth-century compilation [Yangshan] Gushi wenfang xiaoshuo [陽山]顧氏文房小說; the first two of these include notes indicating that the texts were critical editions of Song copies in the possession of the Changzhou 長州 branch of the Gu family (GWX, 2:213–15, 2:186–87, and 1.101–3). The same compilation includes partial editions of the collections Boyi zhi and Jiyi ji, both of which are said to be reprints of Song editions in the Yangshan Gu family’s possession (GWX, 2:216–25 and 2:226–34). A partial Ming dynasty wood-block print edition of Xuanguai lu 玄怪錄 that includes tales from Xu Xuanguai lu also survives; the late Ming compendium Baihai 稗海, compiled by Shang Jun 商浚, includes editions of Duyi zhi, Xuanshi zhi (though Li Jianguo suggests that this may have been based


16 Introduction

Instead Tang tales have come down to us today in large part through the late-tenth-century compendium Taiping guangji, one of the three liter-ary compilations assembled in the 970s and 980s at the direction of the second Song emperor, Taizong 太宗 (r. 976–98), as part of a project of organizing the literature of previous ages for imperial perusal.35 Belles-lettres were placed into Wenyuan yinghua 文苑英華, material especially useful for the emperor into Taiping yulan 太平御覽, and informal anec-dotal material—tales—into Taiping guangji. Taiping guangji includes mate-rial dating from the preimperial period up through the decades just before its compilation, but the bulk of its contents are from the Six Dynasties and Tang. The compendium organizes its material into ninety-two topics (some of which contain subtopics), beginning with Daoist hagiographies and proceeding through accounts of magicians, Buddhist monks and devo-tees, gods, ghosts, retribution, predestination, were-beasts, and many other topics.36 Within Taiping guangji, items drawn from a particular collection would thus be scattered throughout any number of the compendium’s topical chapters according to how the editors classified its contents. The organization by topic reflects the motivations behind the Taiping guang-ji ’s compilation in providing information arranged for easy access and is a measure of how its material was regarded in the tenth century. But the process of placing tales into categories was also a process of removing them from the contexts in which they had been received. Though most of the items in Taiping guangji are followed by a short citation indicat-ing the source, some tales lack attributions; in other cases comparison with other sources or citations reveals discrepancies with the Taiping guangji attribution. For most tales, no other sources survive that could give an alternative or corrective to the Taiping guangji information, but from the instances of overlap that we do have it is clear that the compilation is not

on a Song edition itself in part reconstructed from TPGJ), and Youyang zazu (among other works). The Beijing Library holds Ming manuscripts of Duyi zhi and Xuanshi zhi, as well as a Ming printing of Youyang zazu (Xulu, 2:771, 810, 719). Ganze yao was included in several Ming collectanea and Qing printed editions, though these may have been at least partly reconstructed from TPGJ (Xulu, 2:800–801). Qing printings of Sanshui xiaodu 三水小牘 also survive (Xulu, 2:962).35. TPGJ, in five hundred juan, was compiled by an editorial committee of thirteen headed by Li Fang and presented to the throne in 978 (TPGJ, “Memorial on Taiping guangji ” 太平廣記表, I–II).36. For a list of all of the topical categories covered, see Schafer, “Table of Contents.”


Introduction 17

entirely reliable as a means for reconstructing the collections from which it was constructed.37 Equally problematic for the quest to understand the state of Tang tales in their own time is the fact that the Taiping guangji editors appear in at least some cases to have edited the texts of the tales they included, preserving the main narrative of any given tale but removing brief sections at the beginning or (more often) end of the tale in which the recorder comments on his sources and/or his motivations for recording his tale.38 The pattern is prevalent enough to suggest that the editors made a systematic (though not universal) practice of removing such information, perhaps on the grounds that it was extraneous to the narrative. Again, the occasional survival of tales and collections in other texts—such as the Song or Ming printings mentioned above, as well as the occasional tale found also in Wenyuan yinghua or a writer’s literary collection—preserves at least some of that information, but raises the question of what else has been lost. Compounding these difficulties is the rocky course through which Taiping guangji itself comes down to us. The compilation was completed in 978 and printed three years later in 981, but the printing blocks were later withdrawn because its contents were considered inessential for younger scholars. It is not known how many copies were printed at the time, though it is generally assumed that the number was small.39 At least a few copies of the book seem to have survived in public circulation through the Song: three early Southern Song bibliographies list copies of the book; and the Zuiweng tanlu, a work that both discusses storytelling and reproduces a number of stories said to be in the repertoire of professional storytellers, notes in its opening sentences that the art of storytelling (xiaoshuo) required,

37. For a discussion of this and other problems with the works cited in TPGJ, see Ye Qingbing, “You guan Taiping guangji,” 11–43.38. See chapter 1, note 10, for examples, as well as Li Jianguo’s discussion of the TPGJ compilers’ editorial practice and the types of changes they routinely made in “ ‘Li Wa zhuan.’ ” Li Jianguo makes the point that the compilers’ primary interest was in the stories 故事 recounted in the works they selected rather than in authorship or wording, leading them to take considerable freedom with the texts.39. See the entry on Taiping yulan and Taiping guangji in Wang Yinglin 王應麟, Yuhai 玉海, 54.34a–35a, as well as Zhang Guofeng’s discussion of TPGJ ’s circulation in the Northern and Southern Song (“Taiping guangji ” banben kaoshu, 6–10). Wang Shaoying’s 汪紹楹 “Note on Punctuating and Collating” 點校說明 (TPGJ [1961], i–ii) and Wang Guoliang’s 王國良 “Brief Account of Taiping guangji ” 太平光記概述 (TPGJ [1987], iii) also discuss Taiping guangji ’s early history. Ye Qingbing suggests that in fact no copies were printed (“You guan Taiping guangji,” 15).


18 Introduction

“in youth, familiarity with Taiping guangji and, in adulthood, study of the histories through the ages” 幼習太平廣記，長攻歷代史書.40 But by the time renewed interest in Taiping guangji led to a series of reprintings in the sixteenth century, Song copies were scarce. The Ming dynasty edition that was used as the basis for the standard modern printed edition of the compilation was itself produced from a copy that, in the words of its editor, Tan Kai 談愷, “had been copied and recopied over a long period of time and was full of wrong characters to the point that I couldn’t even punctuate it” 傳寫已久。亥豕魯魚。甚至不能以句. Tan Kai and a group of friends then spent considerable time collating and correcting the text, which was first printed in the 1560s.41 Though other editions (both print and manu-script) whose editors worked at least in part from Song editions survive, the discrepancies among all the editions, in the wording of tales and even in attributions, are significant.42

Beyond the particular problems of Taiping guangji and transmission of tales beyond the Tang, however, in studying eighth- and ninth-century tales we are dealing with products of an age of manuscripts, during which precise reproduction of an authored text was not always valued, or expected, in the way it is today, especially for texts by contemporary and not-yet-canonical writers. The alterations and deletions evidently introduced by the Taiping guang ji editors are just one step in a continuous process of bringing variants into texts, whether an accidental substitution of one word for another or the skipping of a line, or a conscious decision to change a text.43 As Bernard Cerquiglini has said in reference to Western medieval textual practices, “medieval writing does not produce variants; it is variance.”44 Several recent books and articles have called new attention to the impact of manuscript

40. Wang Guoliang, “Brief Account,” in TPGJ [1987], iv; Luo Ye, Zuiweng tanlu, 3.41. Tan Kai’s preface is dated 1566 (TPGJ, ii); Zhang Guofeng’s discussion suggests that the book was printed in the same year (“Taiping guangji ” banben kaoshu, 25), but Wang Guoliang dates it to 1567 (“Brief Account,” in TPGJ [1987], iv).42. Zhang Guofeng, “Taiping guangji ” banben kaoshu, 14–73, gives detailed descriptions of nine surviving pre-twentieth-century editions of TPGJ, four of which appear to have been based on or collated with Song editions.43. For a more detailed discussion of textual variance in the texts of Tang tales, see Allen, “Tang Stories,” chap. 2. Christopher Nugent also gives a thorough catalogue of the types of lexical variants found across the Dunhuang manuscripts of Wei Zhuang’s 韋莊 long narrative poem “Lament of the Lady of Qin” 秦婦吟, which are representative of the vari-ants in the texts of tales as well; see Manifest in Words, 51–54 and 297–309.44. Cerquiglini, In Praise of the Variant, 77–78; emphasis in original.


Introduction 19

practices on the preservation of poetic works in medieval China. Xiaofei Tian and Christopher Nugent have both demonstrated that manuscript reproduction means fluidity: that virtually every oral or written perfor-mance of a poem introduces change. Tian writes: “To the degree that [scribes] were engaged in the production of manuscript copies by copying, editing, altering, and revising, we are no longer talking about the readers’ reception of a stable text, but about the readers’ dynamic participation in the very process of creating a text that is essentially protean. In such a para-digm, the author is still important, but the author no longer occupies the stable central position as an all-powerful and controlling presence in rela-tion to his or her work.”45 Nugent, speaking specifically of the Tang, notes that “even when a copyist, whether he was a professional scribe or simply a reader who wanted his own copy of a poem, may have intended to stay fully true to the original, changes inevitably crept in.”46 Both emphasize that any instance of reproducing a text was an opportunity for textual variance, whether through unintentional miscopying or a conscious decision to alter the text. No less than in poetry and perhaps more given the tale’s relative marginality, such variance was endemic in the material I examine here. Whenever we have more than one text of a given tale, lexical discrepan-cies are inevitable. The Taiping guangji and Wenyuan yinghua editions of Shen Jiji’s “Inside the Pillow” 枕中記, for example, give a vivid illustration of how much variance might be introduced into different texts that still unquestionably represent the same tale (a topic I return to in chapter 6). “Inside the Pillow” is an extreme example, with variance in nearly every sentence, but when two texts exist for a given tale they are never identical. All such instances and many others speak to the active involvement of transcribers and editors in making (remaking) texts of a tale. As Campany notes of Six Dynasties anomaly accounts, “Having been written on paper or silk . . . stories continued to be socially produced in a literal sense . . . in ways that go beyond the reception of a fixed work to include the alteration of the text itself.”47 There is no authoritative text, but rather the contingent productions of multiple hands, beginning for a written text at the point when it is first recorded and extending through multiple transcriptions

45. Tian, Tao Yuanming, 8–9.46. Nugent, Manifest in Words, 21.47. Campany, Signs, 26.


20 Introduction

and editions and compilations, perhaps even including revision by that original recorder himself. Any two documents recording the same tale were far more likely to be at least slightly different than to be identical. The texts we have today thus represent the accretion of change over centuries. The textual difficulties that Taiping guangji presents are enough to tell us that editorial alteration of (or carelessness with) the texts of tales remained standard practice long after a more scholarly approach to Tang poetry led to the preservation of variants as alternative readings even when an editor or reader-scribe selected one of those variants for inclusion in the main text. For tales, we have neither the multiple Dunhuang manuscripts that allow Nugent to study lexical variants in documents of the same poem produced within a few decades of each other nor the rich array of alterna-tive readings preserved within received editions of Tao Qian’s 陶潛 works that Tian studies. In the absence of actual Tang documents for tales, the variance in the received editions give us only a general understanding of how specific texts were treated in their own day. For that reason this book assumes rather than trying to demonstrate that lexical fluidity was part and parcel of the condition of being a text in medieval China, a necessity born of the nature of our sources.

This Book

Instead I explore a different order of instability and fluidity, namely, the larger-scale variations in the ways a given set of events or a given narra-tive formula is put to use in different tales. I draw on both tales originally recorded within collections and individually circulating tales, focusing on material from collections of longer narratives (i.e., “accounts,” including both zhiguai and chuanqi) rather than the shorter notes on court and capi-tal gossip found in what today are generally called “notes” or biji 筆記 (the material often called xiaoshuo in the Tang).48 I use chiefly material dated to a period beginning with the last decades of the eighth century and extend-ing through the first decade of the tenth century. Few Tang tales or collec-tions come with dates attached, making precise dating impossible, but

48. Since there is sometimes overlap in the events recorded in tale collections and biji, however, I do occasionally refer to those collections.


Introduction 21

the greatest concentration of the material I consider was likely recorded between the 780s and the 860s. Far more survives from the late eighth and ninth centuries than from the first half of the dynasty or the very end, and it appears that collecting and recording tales of unusual events saw an upsurge in popularity in the early ninth century that lasted through at least the middle of the century. At the same time, however, more documents of every sort have survived from the second half of the dynasty, and it is possi-ble that this perceived surge is no more than a function of what happens to be preserved.49

Despite this strong bias toward tales from the second half of the dynasty (a few of the tales I reference may even have been recorded after its fall), I continue to use the dynastic label “Tang” to describe the material in the pages that follow. This is in part because the material has conventionally been described this way in scholarship to date—literary historians writing in English or Chinese speak of “Tang dynasty tales,” or Tangdai xiaoshuo 唐代小說, and “Tang chuanqi,” or Tangren chuanqi 唐人傳奇, and other simi-lar variations. But more significantly, it is also because the material itself is overwhelmingly directed at recounting events from that period. The vast majority of tales in the corpus deal with incidents said to have occurred in the period from the dynasty’s founding in the first decades of the seventh century through the increasingly chaotic years of the last decades of the ninth century. The corpus as a whole thus concerns the Tang as an integral era, and the strong focus on that era suggests that the men who recorded these tales also saw themselves as documenting events of their dynasty.50

In approaching this material, I have attempted to take up Dudbridge’s challenge to “reach beyond the inherited categories of one or another gener-ation of anthologists or literary historians . . . [and] confront the primary texts as best we can in their own environment and accept all the complexity that may face us there.”51 This means looking beyond the standard anthol-ogy pieces that have emerged as the Tang tale canon in the course of the past few centuries to examine tales that have fared less well in the test of

49. On the uneven survival of materials from different periods within the Tang, see Twitchett, Writing of Official History, 202–5.50. In fact, the tales we have often recount events dated to several decades before the tales themselves are believed to have been recorded, suggesting a kinship felt with men of previ-ous generations.51. Dudbridge, “Question of Classification,” 197–98.


22 Introduction

time, whether because they are less accessible to later readers, less interest-ing, or simply by chance. Looking at this broader swath of material allows us to notice aspects that would have been obvious to the contemporary reader—for example, how certain tales gain in tension and complexity by violating expectations or how others present novel renderings of a conven-tional theme. It also enables us to better understand how these tales func-tioned in their own time. It is only when we look at many tales side by side (only a portion, we must assume, of what once existed) that we see how this recorded material drew on and existed as gossip and lore, forming a sort of public narrative property on which different writers (and presumably storytellers) could draw at will. But the inclusive perspective also enables us to see the considerable range of different uses to which this shared mate-rial was put. However much they rely on received information and familiar narrative formulae, many written tales are not merely records of hearsay recorded to preserve information as it was received. Rather they are finely wrought texts in which writers explore the significance of the scenarios presented in an inherited theme or set of events. Confronting this body of works thus also means looking beyond the assumptions about fictionality, authorship, and authoritative texts that have formed around these texts in much modern scholarship and that we take for granted in our world of texts fixed in print and strictly regulated copyright. When we treat the standard anthology pieces in isolation, it is easy to perceive in them an attention to detail and psychology and a care-ful sense of pacing and plot that look very much like the result of inten-tional fiction making. When we read more broadly, however, we see that the anthology pieces belong to a corpus whose motivations and origins are very different from those we associate with fiction today, only in part because of its reliance on received information. The written tale’s author-ity in the eyes (and hands) of readers was also correspondingly weak: one Tang reader might have regarded a tale as an authored text whose writer had claim to the story he told and the words in which he told it, but with equal ease another might treat it as narrative material available for further rewriting to suit his own needs. The vision of these tales that results from reading more of them and paying attention to the intricate web of interre-lationships among them is more medieval, more complex, and ultimately more interesting than the picture constructed from a small sampling.

h


Introduction 23

In what follows I begin by situating Tang tales in the context of the exchange of stories among elite men, which I argue lies behind the writ-ten tales that survive today.52 The combination of shared information and particular expression found in the tale corpus suggests that written tales and the practice of writing tales grew out of items of gossip and hearsay circulating among the men who were both the recorders and the first audi-ence for the tales. But the diversity of forms that a single item of hear-say may take within a cluster of tales that all develop from the same topic attests to the fact that, for many writers, close fidelity to the source per se was neither a goal nor a value; rather, the source was a starting point from which the writer developed his own tale. Chapter 2 looks at tales that relay a particular type of information: gossip about public figures. Whatever stories they tell, these tales fasci-nate in part because of who they are about. Tales about emperors and high officials are particularly revealing because we are able to reconstruct some of the historical context that lies behind the tales’ presentation of their protagonists and those protagonists’ deeds. We see that, like any gossip, these tales not only inform and entertain their audiences, but also guide those audiences toward a particular understanding of historical events. When the tales tell a different story from that found in the formal histor-ical record, we realize that the meaning of events recorded so matter-of-factly in the histories was once a matter of debate. Tales offer us a glimpse of a world in which recent history is more contested than it appears in the

52. A word on terminology: Throughout this book I use the English word “tale” as an imperfect designation for the written accounts that are my focus. A “tale” in this termi-nology refers to a particular telling of a set of events. A tale recounts a “story,” which is a narrated set of events (not associated with any particular telling). Two or more tales can be considered to tell the same story if they center on the same core occurrences, even if in other details they may be very disparate. Using the language of twentieth-century narra-tology, the tale is the particular “discourse” through which the story is conveyed, the “récit” to the story’s “histoire” or the “sjuzet” to its “fabula.” A tale therefore implies a particular “text” in which the tale is written—defined by relatively fixed narrative order and word-ing—though my usage allows for some variance in wording within different “editions” of the same tale; the differences between the different editions in which a tale may survive reflect that inevitable variance. Thus the same story may be told in two different tales, and each of those tales may survive, with textual variants, in more than one edition, for exam-ple, both in TPGJ and in WYYH or a writer’s literary collection. (For standard discus-sions of the distinction between story and discourse, see, for example, Genette, Narrative Discourse, 25–28 [including nn. 1–2]; and Culler, Pursuit of Signs, 169–87.)


24 Introduction

received historical record and in which the recorders of tales are keenly interested in, and interested in influencing, the interpretation of recent events. In Chapters 3 and 4 I turn to an exploration of the ways that inherited lore informs and is transformed in tales of the eighth and ninth centuries. I focus on a stock situation, a lone man’s encounter with a stranger with a secret, that is found repeatedly in tales of this period and that by the late eighth century almost invariably took the shape of a particular, even formulaic, sequence of events. I outline this formula and examine both its typical operations and a selection of tales, including some of the best-known today, that question the assumptions about the separation between human and other that lie at the heart of the formula. I then turn to tales that use the same lore-based formula as a stage for contrivance and wit. In contrast to the great majority of the corpus, such tales call attention to the artificiality of the stories they tell. The popularity of witty tales in the mid-ninth century suggests a new attitude toward tales on the part of both writers and readers in which the writer’s inventiveness took center stage. Chapters 5 and 6 investigate changing attitudes toward authorship and the texts of tales in the eighth and ninth centuries. For most ninth-century readers, the texts of tales remained open. Written tales did not yet have authority as exclusive versions of the stories they told; a reader who wanted to possess a copy of a story he had read himself could copy it fairly faith-fully, or he could make changes big and small to suit his own needs without acknowledging his source. My focus is not the small changes in wording that are inherent in medieval manuscript transmission, but rather the wholesale rewriting of tales that were already in circulation. Even when working from a previous written text, most story collectors would make changes. However, a more textual and author-centered attitude began to emerge in the mid- to late ninth century. This shift toward regarding tales as potentially textual, authored works represented a significant change in the perception of writ-ten stories that moved them for the first time into the realm of belletristic writing. Though never the normative way of reading tales in the Tang, it was a first step in the process through which these tales, disparate in origins and treatment, came to be seen as a cohesive corpus with a recognizable canon.


introduction - harvard university...introduction 3 with information about how the world around him...

Documents